On Mon, May 19, 2003 at 02:13:58PM -0600, scott.marlowe wrote:
> Since shared_buffers are fixed and don't change during operation, they
> don't have the same danger that sort_mem does of dynamically running the
> machine into the ground should too much memory gets allocated to sorts.
> sort_mem is a limit per sort, so theorectically, a complex query could
> generate more than one sort, and a handful of clients running large sorts
> could run the machine of out RAM and into a swap storm as tries to service
> all the backend process sorts. So, while having sort_mem too small costs
> a little in performance, having it set too high can result in your server
> coming to a crawl under load, which is definitely worse.
Hmm... this could explain some of the memory alloc errors people have
been posting about. It would be very useful if pgsql could limit the
amount of memory used by a connection, or better yet, used across all
connections. This way you could ensure that you never start swapping.
--
Jim C. Nasby (aka Decibel!) jim@nasby.net
Member: Triangle Fraternity, Sports Car Club of America
Give your computer some brain candy! www.distributed.net Team #1828
Windows: "Where do you want to go today?"
Linux: "Where do you want to go tomorrow?"
FreeBSD: "Are you guys coming, or what?"